Reliability
Retries and dead-letter queue
Retries re-attempt a failed step after a delay, and a dead-letter queue holds the items that still fail after the last retry so a person can inspect and replay them.
How it works in practice
Many failures are temporary: a timeout, a rate limit, a service restarting. Retrying with a growing delay, called exponential backoff, clears most of them without anyone noticing. Some failures are permanent, such as a missing customer or a rejected field, and retrying those forever wastes time and can hide the problem.
A dead-letter queue is where those items go after the final attempt. Each one keeps its payload and error, someone is told, and once the cause is fixed the items can be replayed. Combined with idempotent steps, this means a failure delays work without losing it or doing it twice.
A worked example
An order sync to the warehouse retries up to five times over 30 minutes. Orders that still fail, usually for an unknown SKU, land in a dead-letter table shown on the operations dashboard, with a replay button once the SKU is added.
Examples in this glossary describe typical builds for companies of 1 to 1,000 people. They are not named client results.















